ReViNE: Reallocation of Virtual Network Embedding to Eliminate Substrate Bottleneck

ReViNE: Reallocation of Virtual Network Embedding to Eliminate Substrate Bottleneck Shihabur R. Chowdhury, Reaz Ahmed, Nashid Shahriar, Aimal Khan, Raouf Boutaba Jeebak Mitra, Liu Liu

Virtual Network Embedding (VNE) 10 a 10 12 e 80 15 55 A B c b c 10 10 10 20 d 5 5 e f 20 20 a C 60 22 15 12 90 50 20 10 D E F d 17 85 f G 17 H 70 65 b 25 2

Virtual Network Embedding (VNE) 10 a 10 12 100 15 55 A B c b c 10 10 10 20 d 5 5 a C 60 27 20 12 110 50 20 10 D E F 85 22 25 e f 20 20 G 22 H 70 85 b 3

Impact of Dynamicity: An Empirical Study 300+ SNodes, 900+ SLinks. (AS6461), 4 8 VNodes/VN (50% conn. pr.) Poisson Arrival (10VNs/100 T.U.), Exponential Lifetime (1000 T.U.) Optimal embedding that minimizes total bandwidth consumption 4

Impact of Dynamicity: An Empirical Study Acceptance Ratio capped at ~50% 300+ SNodes, 900+ SLinks. (AS6461), 4 8 VNodes/VN (50% conn. pr.) Poisson Arrival (10VNs/100 T.U.), Exponential Lifetime (1000 T.U.) Optimal embedding that minimizes total bandwidth consumption 5

Impact of Dynamicity: An Empirical Study ~30% links utilized >= 70% ~40% links utilized <= 10% 6

Impact of Dynamicity: An Empirical Study ~30% links utilized >= 70% ~40% links utilized <= 10% Skewed Substrate Link Utilization impacts Acceptance Ratio!! 7

Key Question: How to cope with the dynamicity in Network Virtualization when little or no information about the future is available? 8

(One Possible) Answer: Periodically adjust the embedding to eliminate bottlenecks and optimize resource usage 9

The Problem Reallocation of Virtual Network Embedding (ReViNE) Given a Substrate Network and a set of embedded Virtual Networks 10

The Problem Reallocation of Virtual Network Embedding (ReViNE) Given a Substrate Network and a set of embedded Virtual Networks Migrate Virtual Nodes to New Substrate Nodes 11

The Problem Reallocation of Virtual Network Embedding (ReViNE) Given a Substrate Network and a set of embedded Virtual Networks Migrate Virtual Nodes to New Substrate Nodes Migrate Virtual Links to New Substrate Paths Objective: Eliminate Substrate Bottlenecks* and Minimize Resource Usage** * Links with utilization >= % ** In our case, bandwidth consumed by virtual links 13

Our Proposal A suit of solutions to ReViNE ReViNE-OPT ILP-based optimal solution* (NP-Hard) ReViNE-FAST Simulated Annealing-based heuristic * Details is in the paper 14

Do We Need A Heuristic? Computing Optimal Solution is Very Expensive H/W Configuration: 8x10 Core Intel Xeon E5 CPU, 1TB RAM Observed limits for ILP: 50 100 Node SN with < 60VNs took several hours and several 10s of GB RAM ILP Can Yield Impractical Solutions A practical solution contains a sequence of operations to reach the re-optimized state (also satisfy make-before-break constraint) Not possible to model in ILP. Final state obtained from ILP can be unreachable without violating make-before-break constraint. 15

Do We Need A Heuristic? Computing Optimal Solution is Very Expensive H/W Configuration: 8x10 Core Intel Xeon E5 CPU, 1TB RAM Observed limits: 50 100 Node SN with < 60VNs took several hours and several 10s of GB RAM ILP Can Yield Impractical Solutions A practical solution contains a sequence of operations to reach the re-optimized state (also satisfy make-before-break constraint) Hard to model in ILP. Final state obtained from ILP can be unreachable without violating make-before-break constraint. 17

Heuristic Design Our Objectives are Conflicting Minimize Bottleneck Links Minimize Bandwidth Usage 18

Heuristic Design Our Objectives are Conflicting Minimize Bottleneck Links Minimize Bandwidth Usage Distribute load across substrate links Paths can become longer 19

Heuristic Design Our Objectives are Conflicting Minimize Bottleneck Links Distribute load across substrate links Paths can become longer Minimize Bandwidth Usage Route Virtual Links on Shorter Paths Substrate links on shorter paths can become bottlenecks Instead of an one-shot algorithm, use a meta-heuristic (Simulated Annealing) to explore the solution space and find a balance. 22

Simulated Annealing: Neighborhood Generation Bottleneck Substrate Link Reconfiguration Select a bottleneck substrate link and reroute virtual links using that bottleneck link until it is no longer a bottleneck. Virtual Node Migration Randomly select a VN and re-embed a random virtual node and incident virtual links. Virtual Link Migration Randomly select a VN and reroute a randomly selected virtual link. 23

Exploiting Multi-core CPU Parallel Simulated Annealing Searches Seed Solution-0 Initial State Seed Solution-1 Seed Solution-k 26

Exploiting Multi-core CPU Parallel Simulated Annealing Searches Seed Solution-0 Search Thread 0 (CPU0) Initial State Seed Solution-1 Search Thread 1 (CPU1) Seed Solution-k Search Thread k (CPUk) 27

Exploiting Multi-core CPU Parallel Simulated Annealing Searches Seed Solution-0 Search Thread 0 (CPU0) Initial State Seed Solution-1 Search Thread 1 (CPU1) Best Solution from All Searches Seed Solution-k Search Thread k (CPUk) 28

Evaluation: Setup ReViNE-FAST compared with ReViNE-OPT and SA-realloc* Parameters 50 100 node synthetic substrate network Larger test cases with 1000 node (heuristic only comparison) Mean degree between 3.6 4 Mean substrate link utilization 60% - 80% Bottleneck substrate link threshold 70% - 90% * Masti, S,. et al. Simulated Annealing Algorithm for Virtual Network Reconfiguration, 8 th Euro-NGI Conference on Next Generation Internet, IEEE, 2012, pp. 95-102. 29

ReViNE-FAST Performance Highlights Within ~19% of optimal (ReViNE-OPT) on avg. ~3x less cost compared to SA-realloc on avg. ~5% more VNs accepted on avg. when combined with optimal VN embedding algorithm 30

Summary ReViNE is one possible way to address the dynamicity in VN arrival/departure ReViNE-FAST, a simulated annealing based heuristic performs ~19% within the optimal (empirically evaluated) ReViNE-FAST performs ~3x better than S.O.A Simulated Annealing-based heuristic 31

Questions? Source Code CPLEX: https://github.com/srcvirus/vne-reallocation-cplex Simulated Annealing: https://github.com/srcvirus/vne-reallocation-sa 32

Backup 33

ReViNE-FAST vs ReViNE-OPT 34

Impact of Reallocation 35

ReViNE-FAST vs SA-Realloc (Large Cases) 36

ReViNE-FAST Convergence (Large Cases) 37

State-of-the-art Reactive One-shot Approaches: Reallocate VNs when a new VN cannot be embedded [1][2] Proactive One-shot Approaches: Periodically reallocate VNs [3][4] Meta-heuristic Approaches: Simulated Annealing [5], Particle Swarm Optimization [6] [1] Y. Zhu et al., Algorithms for assigning substrate network resources to virtual network components, IEEE INFOCOM, 2006. [2] M. Yu, et al. Rethinking virtual network embedding: substrate support for path splitting and migration, ACM SIGCOMM CCR, 38(2),2008, pp. 17 29. [3] N. F. Butt, et al. Topology-awareness and reoptimization mechanism for virtual network embedding, Int. Conf. on Research in Networking 2010. [4] P. N.Tran, et al., Optimal mapping of virtual networks considering reactive reconfiguration, IEEE CloudNet, 2012. [5] S. Masti, et al. Simulated Annealing Algorithm for Virtual Network Reconfiguration, 8 th Euro-NGI Conf. on Next Generation Internet, IEEE, 2012. [6] Y. Yuan, et al., Discrete particle swarm optimization algorithm for virtual network reconfiguration, Int. Conf. in Swarm Intelligence, 2013. 38